Yakubu et al. (2025) Global Bias-Corrected CORDEX Datasets at Half Degree Resolution

Identification

• Journal: Scientific Data
• Year: 2025
• Date: 2025-11-12
• Authors: Fuseini Jacob Yakubu, Jürgen Böhner, Udo Schickhoff, Thomas Scholten, Shabeh ul Hasson
• DOI: 10.1038/s41597-025-06200-4

Research Groups

• HAREME Lab, Institute of Geography, ESRAH, Universität Hamburg, Hamburg, Germany
• Institute of Geography, ESRAH, Universität Hamburg, Hamburg, Germany
• Department of Geosciences, Chair of Soil Science and Geomorphology, University of Tübingen, Tübingen, Germany

Short Summary

This paper presents GloBCORD-HD, a new quasi-global dataset of bias-corrected Coordinated Regional Climate Downscaling Experiments (CORDEX) at 0.5° spatial and daily temporal resolution for historical (1950–2019) and future (2020–2099) periods under three Representative Concentration Pathways (RCPs). Comprehensive validation demonstrates that GloBCORD-HD significantly reduces biases, improves regional extremes representation, and enhances climate signal consistency, enabling robust global impact assessments.

Objective

• To develop a standardized, high-resolution, quasi-global dataset of bias-corrected CORDEX experiments using a uniform method and a common observational dataset to provide consistent climate simulations for robust regional and global climate change impact assessments.

Study Configuration

• Spatial Scale: Quasi-global (excluding polar regions), 0.5° spatial resolution.
• Temporal Scale: Daily temporal resolution. Historical period: 1950–2019 (or 1960–2019 for MPI-LR_CDX). Future period: 2020–2099 for RCP2.6, RCP4.5, and RCP8.5 scenarios.

Methodology and Data

• Models used:
  • Dynamically downscaled CORDEX experiments (Phase 1, CMIP5) driven by three Earth System Models (ESMs): MPI-M-MPI-ESM-LR, ICHEC-EC-EARTH, and NOAA-GFDL-GFDL-ESM2M.
  • Regional Climate Models (RCMs) involved in CORDEX downscaling: RCA4, REMO2009, CCLM4–8–17-CLM3-5, CRCM5, WRF, and CCLM5-0-2.
  • Bias correction algorithm: ISIMIP3BASD v2.5 (a trend-preserving quantile mapping approach).
  • Software tools: Climate Data Operators (CDO) version 2.4.1 for regridding and merging, ibicus Python package for bias correction.
• Data sources:
  • Observational dataset for bias correction: GSWP3-W5E5 v2.0 (daily, 0.5° spatial resolution, 1901–2019).
  • Raw CORDEX 12 experiments obtained from the Earth System Grid Federation (ESGF) metagrid portal.
  • Variables: Daily mean near-surface air temperature (tas), daily minimum near-surface air temperature (tasmin), daily maximum near-surface air temperature (tasmax), daily total precipitation (pr), daily mean surface downwelling longwave radiation (rlds), daily mean near-surface relative humidity (hurs), and daily mean near-surface surface winds (sfcWind).

Main Results

• GloBCORD-HD datasets successfully reduce historical biases across all seven climate variables. For example, temperature biases were reduced to within ±0.5 °C, and precipitation biases (initially >50 mm/day in some regions) were reduced to within ±1 mm/day.
• The bias correction improves the representation of regional extremes and enhances the consistency of climate signals.
• Spatial validation shows effective bias reduction across different geographical regions and climate zones, with boundary discontinuities largely excluded or smoothed out (NRMSD < 0.5 for most variables, "Good" for precipitation).
• Temporal validation demonstrates strong alignment of bias-corrected historical simulations with observations and consistent future projections, with adjustments to mean biases in future projections reflecting historical bias structures.
• Statistical evaluation (mean bias, standard deviation, Kolmogorov-Smirnov statistics) confirms significant improvements in overall accuracy and distributional agreement for temperature and precipitation.
• The extension of the historical period (2006–2019) using RCP8.5 data showed negligible impact on model-observation agreement.

Contributions

• Provides the first quasi-global, bias-corrected, dynamically downscaled CORDEX dataset at 0.5° resolution, addressing a critical gap for consistent impact assessments.
• Utilizes a consistent and sophisticated bias-correction methodology (ISIMIP3BASD v2.5) against a single, widely adopted observational dataset (GSWP3-W5E5 v2.0), ensuring uniformity for impact assessment communities like ISIMIP.
• Leverages the strengths of dynamically downscaled Regional Climate Models (RCMs) to explicitly resolve regional physical processes, topographic effects, and land-surface interactions, offering a superior representation of local climate drivers and extreme events compared to statistically downscaled Global Climate Model (GCM) products.
• Offers a comprehensive uncertainty matrix by combining three RCP scenarios (2.6, 4.5, 8.5) from three Earth System Models (ESMs) with varying Equilibrium Climate Sensitivities, providing a robust foundation for climate risk assessment.
• Reduces CORDEX 12 domain overlaps, decreasing overall data volume and providing seamless, consistent input data for robust climate impact assessments.

Funding

• Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy- EXC 2037 “CLICCS-Climate, Climatic Change, and Society”-Project Number: 390683824.
• DFG-funded project “Sensitivity and Response of Treeline Ecotones in the Nepal Himalaya to Climate Warming - TREELINE-II” (DFG, BO 1333/4-1).
• Open Access Publication Fund of Universität Hamburg.
• World Climate Research Programme (WCRP) for CORDEX datasets.
• Projekt DEAL (for Open Access funding).

Citation

@article{Yakubu2025Global,
  author = {Yakubu, Fuseini Jacob and Böhner, Jürgen and Schickhoff, Udo and Scholten, Thomas and Hasson, Shabeh ul},
  title = {Global Bias-Corrected CORDEX Datasets at Half Degree Resolution},
  journal = {Scientific Data},
  year = {2025},
  doi = {10.1038/s41597-025-06200-4},
  url = {https://doi.org/10.1038/s41597-025-06200-4}
}